Update Classification

Confusion matrix 1

Confusion matrix 2

Statitics

# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary           0.9
2 kap      binary           0.8
3 j_index  binary           0.8

From these two matrices we can see that there might be obvious results. I still have some difficulties interpreting these matrices and need to look more deeply into this. Actually I thought these matrices are not suitable for my research, but I think we have to inlude them anyways to get the full score for this assignment. Maybe one of my peers know what I can conclude from them? The next tabs shows a very interesting graph that might also help for this.

Update: Decision trees

Update: Feature Selection

# A tibble: 3 x 3
  .metric  .estimator .estimate
  <chr>    <chr>          <dbl>
1 accuracy binary         0.7  
2 kap      binary         0.400
3 j_index  binary         0.400


When looking at the features in the model by using the “coef” method I could see that besides the two intercepts of the two playlists all other features are also very similar. I still need to look more deeply into this to see if I can conclude anything from these analyses. The decision tree and random forrests also included interesting results and I need to look more deeply into but thought it might be interesting to include them to my update.

Using the code provided to us for the Feature Selection I think I created this very good graph showing that there is a big difference in the two timbre coefficients for German Hip Hop and Reggeaton. While, as in all the graphs I included in this portfolio before, German Hip Hop’s timbre coefficients are much more distributed than the ones for Reggeaton. This again concludes that there is a typical feature in Reggeaton music that might be a reason for its popularity.

Comparing the two second and first matrix including the predicyted

What role does valence play?


Here is the first visualisation I created for my research. It compares the valence on the x-axis to the energy of the song on the y-axis.From the first impression you might not really see a big difference between the German and the Reggeaton songs. But looking closer, you can see that there is more variance within the German songs than within the Reggeaton songs. As I described in my own expectations, Reggeaton songs sound very positive and as the visualisation shows, German songs vary regarding valence while the Reggeaton songs almost all score very high on it. What we can also see is that both genres are quite energetic although, I think, this energy might be different for both of them. The energy in German songs seems more aggresive to me than the energy in the Reggeaton songs. This might also be visualised in the graph, as the German songs are also destributed more central than the Reggeaton songs. Overall from this result, I would describe Reggeaton songs as more positive and more energetic than German Hip Hop songs. This positivity and energy might also be one of the reasons why Reggeaton is so popular among many people all aroung the world. Its positive and energetic vibe spreads a good mood and the language of the songs do not seem to matter in this regard. German Hip Hop songs on the otherhand might also be very energetic but in different way. Looking at the speechiness in the songs, which i perceive to be very high in German songs could be an interesting factor.

Is variation the key to success?

German song “Mios mit Bars” by Luciano

Reggeaton song “En la Cama” by Nicky Jam feat. Daddy Yankee


I created these two Chromagrams choosing the first song on each playlist. We can see that the first song “Mios mit Bars” by Luciano from the German Hip Hop Playlist is mostly played in A and not that much in G#, G and F#. Additionally, we can see that in the beginning and the end of the measured time of the Chromagram for a lot of notes the magnitude is very low. When looking at the second Chromagram visualising the song “En la Cama” by Nicky Jam featuring Daddy Yankee we can see that the magnitude is higher for almost all notes. Also, the ending and beginning measured for this song’s chromgram do not exhibit such striking lower magnitude as the German song. From these visualisations we might therefore again conclude that this variety of notes and therefore sounds in the Reggeaton song might lead to more popularity among listeners from all around the world than the more monotony of the German song.

More disrtibution for timbre in German Hip-Hop songs than in Reggeaton songs

German Hip Hop outlier - Wenn der Vorhang fällt (feat.Wasi) by Freundeskreis

Reggeaton outlier - No Le Temas a el Beat by Trebol Clan and Hector & Tito


For the cepstograms and the following self-similarity matrices I chose the two outliers of the two playlists.

When comparing the cepstograms regarding timbre we can see that the magnitude for the Reggeaton song is much less distributed than the magnitude for the German song, although the German song highlights c03. As these two are outliers for the actual genres we can again see that variety in tone and in timbre, as it is not the case for the Reggeaton outlier, might indeed a characteristic for Reggeaton. To be sure of that, I want to make further investigations on that and possibly also create cepstograms for typical Reggeaton songs.

Differences also between outliers of the two genres

Wenn der Vorhang fällt (feat.Wasi) by Freundeskreis

No Le Temas a el Beat by Trebol Clan and Hector & Tito


The self-similarity matrices that also portray the two outliers are very diverse as well. Although these two are outliers for each genre we can see that there is a clear difference between the two. To find a more clear and exact interpretation for these, I would like to create more self-similarity matrices with different songs from each playlist.

Update 1: Keygrams

Keygram of “Ginza-Remix” by J Balvin

Keygram of “Die Abrechnung” by Eko Fresh & German Dream Allstars


Comparing these two Keygrams of “Ginza-Remix” by J Balvin and “Die Abrechnung” by Eko Fresh & German Dream Allstars we can see that there are some differences between the two songs. I chose these two songs randomly from the playlists to present typical songs of each genre. I will interpret the Keygrams together with the Chordograms on the following tab.

Update 2: Chordograms

Chordogram of “Ginza-Remix” by J Balvin

Chordogram of “Die Abrechnung” by Eko Fresh & German Dream Allstars


Observing these two Keygrams and the previous Chordograms we can see that for the Reggeaton song “Ginza” by J Balvin there is an emphazise in the last seconds of the song. Also Eb major is stressed in both the Chordo- and the Keygram. It also seems like every 100 seconds there is emphasis on specific notes. For the German song “Die Abrechnung” we can see differences between the Chordo- and the Keygram which show the complete opposite: As the beginning of the song in the Chordogram is stressed it is not emphasized at all for the Keygram. There also seems to be a pattern, but a less symmetrical one than for the Reggeaton song. This could again be a sign for the specific beat in Reggeaton songs causing a repetition.

I want to improve and concentrate a bit more on these two graphs to find maybe more relevant results.

Update 3: Track-Level summaries

Comparison of the two genres

Comparing average timbre coefficients in German Hip-Hop and Reggeaton


This graph is a track-level summary comparing our two genres Reggeaton and German Hip-Hop especially looking at tempo. Additionally features like volume, duration and genre are mentioned in the legend and presented in the graph. Looking at it more deeply we can see that as all other graphs we analysed until now in this portfolio, German Hip-hop exhibits much more variation than Reggeaton does. Although I would have expected the mean tempo to be much higher for both genres, we can see that Reggeaton songs are all positioned in the lower left part of the graph while the German songs are more distributed towards the middle but having there main spot at the same part as the Reggeaton songs. For volume we can see that both genres have a very similar loudness, which I expected to be higher. Therefore, this graph nicely underlines my own perception of the two genres: German Hip-Hop are quite similar in terms of tempo and volume but German Hip-Hop varies more, while Reggeaton songs feature a specific, characteristic tempo.This could again be another reason for Reggeatons popularity.

The second graph concentrates on the averages of timbre coefficients in the both genres. We can see that for some coefficients the two genres are similar while for others they are very different. Again some of the coeffcients show that Reggeaton has more variation.

Volume might be unnecessary for this graph in my case?

Update Tempi Histograms

Histogram “Ohne mein Team” by Bonez MC & RAF Camora

Histogram “Ella Y Yo” by Aventura & Don Omar


These two tempi histograms show the new changes in tempo during the two songs “Ohne mein Team” by Bonez MC & RAF Camora and “Ella Y Yo” by Aventura & Don Omar. From the first grasp you would think that these two songs have a very different change in tempo. But when you look more closely, the y-axis ranges from 0 to 40 for the German song and from 0 to 15 for the Reggeaton song. Therefore, the difference is not big and we might conclude that the tempo is not changing very much and is very similar for both songs. One might conclude that the low change in tempo is characteristic for both genres and a similarity between the two.

What can we conclude until now?

What I can conclude from my visualisations until now, is that there might be significant differences between the two music styles, for instance in their level of energy and valence which might be a reason for why Reggeaton is also listened to by people who do not even speak Spanish. It might be the positive, danceable and energetic sound that makes people from all around the world enjoy listening Reggeaton.

I also can conlcude now that from my visualisations I can perceive more variability in tone and timbre for Reggeaton songs than for German Hip Hop songs which might be one other reason for Reggeaton’s popularity.

The visualisations I created showed very interesting results and I want to look further at audio features like danceability, loudness, tempo and speechiness as a perceive those as very characteristic of both music styles. More importantly want to look more deeply into my results and create more sophisticated visualisations.

Additionally, I still want improve the layout of my portfolio and also compare the two genres looking at specific songs.